A MarkTechPost tutorial shows how to audit and rebuild Datalab's OmniParseBench and check how certain its parser rankings are using document-level bootstrapping. It also evaluates Tesseract and Indra OCR on a custom sample.
Kevin Weil, who led OpenAI for Science, has raised $215 million led by Lightspeed for Warp Sciences, valuing it at $865 million. The startup is building an automated US lab for a drug development step that pharma companies have long outsourced to Chinese firms such as WuXi AppTec.
Microsoft Research's Flint-interactive gives AI agents a language for building interactive charts. Readers can click, filter and highlight related content to explore beyond the author's original view.
In Brussels, a new tool called Evactwin simulates how people move during a crisis to spot bottlenecks in advance. The aim is to help authorities anticipate disasters and coordinate rescue efforts.
Google DeepMind has produced a global map that assigns every 10-meter patch of land on Earth to a distinct forest type. It goes beyond simply recording where trees are.
A Harvard study of 718 firms using Jellyfish data finds that coding assistants and agents produce more code without a matching rise in finished work. Part of the reason is the extra load they create later on, especially in code review.